Light trapping, a new approach to spectrum splitting

Light trapping, a new approach to spectrum splitting

ARTICLE IN PRESS Solar Energy Materials & Solar Cells 92 (2008) 1570–1578 Contents lists available at ScienceDirect Solar Energy Materials & Solar C...

2MB Sizes 0 Downloads 40 Views

ARTICLE IN PRESS Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

Contents lists available at ScienceDirect

Solar Energy Materials & Solar Cells journal homepage: www.elsevier.com/locate/solmat

Light trapping, a new approach to spectrum splitting A. Goetzberger, J.C. Goldschmidt, M. Peters , P. Lo¨per Fraunhofer Institute for Solar Energy Systems, Heidenhofstr. 2, 79110 Freiburg, Germany

a r t i c l e in f o

a b s t r a c t

Article history: Received 14 April 2008 Accepted 4 July 2008 Available online 29 August 2008

In this paper, a novel type of light trap is introduced. It enables photovoltaic conversion with separate solar cells optimized for different frequency bands, which are covered by spectrally selective mirrors. Sunlight is coupled into a transparent medium and converted to diffuse radiation by randomization. It is well known that under these conditions light intensity is enhanced by a factor of 2n2 inside the medium. The unidirectional radiation at the aperture is however only n2. Two types of light trap are presented. The first one employs concentrated radiation incident on a small volume light trap. It is shown that efficiency depends on the ratio (solar cell area)/(input area). A more detailed analysis shows that under regular conditions part of the light is absorbed directly after entering the trap which results in higher efficiencies. We show that a trapping efficiency above 90% is obtainable. The second type of trap is based on a photonic structure covering the surface of the medium. It has high angular selectivity and broad spectral transmission. An analysis of efficiency and loss mechanisms is presented. If both types of trap are combined even higher efficiency is possible. Consequences for solar cell technology are investigated. It is further shown that the light trapping principle can also be applied to large area stationary modules. & 2008 Elsevier B.V. All rights reserved.

Keywords: Light trapping Concentration Photon management

1. Introduction 1.1. Background The efficiency of a solar cell consisting of a single semiconductor material is limited theoretically to about 40% [1]. In order to reach higher efficiency, the solar spectrum has to be subdivided and converted by different solar cells with optimized band gaps. The best approach at present is to use multijunction cells stacked above each other such that the top cell converts the shortest wavelength fraction while being transparent to the longer wavelengths that are in turn converted by other solar cells lower in the stack. In concentrating systems, those multijunction cells consist of monolithically grown III–V compounds. Thin film double or triple junction cells have so far only been realized in amorphous-microcrystalline cells but efficiencies are moderate. The multijunction approach has so far yielded the highest efficiencies but has some limitations. To achieve high efficiencies for these cells, the lattice match conditions have to be regarded. Currently, new technologies like lattice mismatched and metamorphic cells [2] are developed though that alters this limitation. Furthermore, between cells tunnel diodes have to be inserted because the cells have opposite polarity. Very stringent is the

requirement that all cells have to produce equal current because the stack represents a series connection. When the spectral conditions of irradiation change current match is negatively affected. If separate individual solar cells could be used to convert the frequency bands more freedom of design and potentially higher efficiency could be expected. A well-known idea to achieve spectrum splitting for separate solar cells is the fluorescent concentrator [3]. Another very old concept to achieve this type of spectrum splitting is to use dichroic mirrors in a concentrating system that focus different parts of the solar spectrum onto separate solar cells [4]. This concept has been getting renewed attention recently [5]. In this paper, a new method is described that permits to use separate solar cells side by side without the restrictions mentioned.

1.2. The basic concept The light trapping concept presented here has the following properties:

 A transparent volume of the trap filled with diffuse radiation.  Solar radiation enters the trap through an input area bordering air.

 Corresponding author. Tel.: +49 761 4588 5475; fax: +49 761 4588 9250.

E-mail address: [email protected] (M. Peters). 0927-0248/$ - see front matter & 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.solmat.2008.07.007

 The radiation is randomized either at entering the trap or by scattering centres or by lambertian reflectors within the trap.

ARTICLE IN PRESS A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

 Separate solar cells with different band gaps are optically  

coupled to the trap. The surfaces of these solar cells are covered with band-pass mirrors that transmit a specific wavelength range and reflect all other light. Surfaces not covered by solar cells are highly reflective.

In the first part of this paper, we describe the basic principle and its application to concentrating systems. Two versions of light traps will be considered. The first is very straightforward and can be applied to existing systems; the second relies on photonic structures requiring more advanced technology. While these concepts are designed for concentrating systems, we will show in the following part how light trapping can also be applied to non-concentrating, stationary modules. The mechanism of the light trap is based on the thermodynamic theory of Yablonovitch [6]. Randomized light within a slab of transparent material of irregular shape and of refractive index n is enhanced in its intensity with respect to the intensity outside in air by a factor of n2. In his paper, Yablonovitch gives a factor 2n2 but he defines light intensity as being bidirectional. Radiation that illuminates the surface of a solar cell is however unidirectional. Therefore, we restrict ourselves to an enhancement by n2 due to diffuse radiation and total internal reflection. The optically coupled solar cells are exposed to this increased intensity. At the input aperture, there is a drop of intensity by n2 vs. air. In an idealized approach, the input aperture is considered without surface (Fresnel) reflection. If necessary, this can be included but surface reflection can be minimized by antireflection measures. We will first consider the situation with fully diffuse light within the trap and then demonstrate that the direction of the incoming light is important and that by its consideration calculated efficiencies are higher.

remainder of the surface is reflective. A similar concept for light confinement in bulk solar cells has been described by Luque [7,8] using micro CPCs, whereas the trap described here is fully transparent. At the bottom of the trap, there are solar cells and mirrors which may be diffuse reflectors. It has to be emphasized that diffusion of the light is essential to the concept. Exactly where the diffusion is accomplished is not important. As Yablonovitch pointed out, a diffusing surface at one point of the trap in combination with multiple reflections results in a nearly isotropic characteristic, even if the characteristic of the diffuse reflector is not lambertian itself [6]. In a light trap with more than one solar cell, the solar cells are always combined with a band-pass mirror at the trap interface that is transparent for the light which is best used by each solar cell and reflective for all other light. As this is always the case, the mirrors were not separately included in the figures. Solar cells in the figures have always to be seen combined with band-pass mirrors or rear mirrors as described in Fig. 11. We now estimate the efficiency of the light trap. In an idealized approach, we assume the radiation to have an isotropic characteristic and be fully uniform within the trap, the mirrors to be perfectly reflective and the solar cells to be perfectly absorbing, while there is no absorption in the trap itself. In this paper, light intensities are generally defined as being unidirectional. Let Iint be the internal radiation intensity, then the loss through the input aperture Ai is IintAi/n2 (the input area Ai is normally somewhat larger than the illuminated area because of tolerances in focussing). The radiation absorbed by the solar cells with area As is IintAs. The trapping efficiency Ztrap is now obtained simply by setting up the energy balance within the trap:

Ztrap ¼

2.1. The basic principle Fig. 1 shows the first approach to the principle of light trapping. Light is first concentrated in the usual manner by lenses or other optical means. The light trap has small openings corresponding to the focal point of the concentrating optics. The

Iint As Iint As þ Iint ðAi =n2 Þ

and

Ztrap ¼ 2. Concentration by geometric optics

1571

1 1 ¼ 1 þ ðAi =As n2 Þ 1 þ ð1=Fn2 Þ

(1)

From Eq. (1), we see that efficiency depends on the ratio F ¼ As/Ai. That means we can control efficiency by adjusting this ratio. In Fig. 2, efficiency is plotted vs. this ratio for n ¼ 1.5 and 2.0. For example, for equal area and n ¼ 1.5, As/Ai ¼ 1: Ztrap ¼ 0.69 and for As/Ai ¼ 2: Ztrap ¼ 0.82. In practice, this means that if we want to achieve high trapping efficiency, we have to concentrate the light first and then deconcentrate it in the light trap. It should also be kept in mind that Eq. (1) applies to every fraction of the solar spectrum separately. Therefore, the total area of solar cells is the product of the number of spectral subdivisions and the area of the solar cells. In other words if, for instance the spectrum is split into three parts, the total cell area is 3As (assuming equal area solar cells). We can also estimate the radiation intensity in the light trap. Qualitatively we can state that the light intensity Iint should drop with increasing As/Ai and efficiency should rise because emission through Ai is lowered. If Iinc is the concentrated radiation impinging on Ai, then without solar cells (for As ¼ 0) Iinc ¼ Iint0/n2. Iinc is the concentrated radiation averaged over the entire input area. When solar cells with area As are added we obtain Ai Iint I þ As Iint ¼ Ai Iinc ¼ Ai int0 n2 n2

Fig. 1. Principle of a light trap. Light is concentrated by lenses or other optical means. The concentrated beam enters the light trap through a small aperture. Inside the trap the light is diffused. Solar cells with optical mirrors are optically coupled to the trap. The optical mirrors reflect all light the solar cell in question cannot use. The trap surface is reflective everywhere apart from the apertures and the solar cells. By that means light is spectrally selective concentrated onto different cells.

Iint 1 ¼ Iint0 1 þ n2 ðAs =Ai Þ

(2)

or Iint 1 1 ¼ ¼ Iinc ð1=n2 Þ þ ðAs =Ai Þ ð1=n2 Þ þ F

(3)

ARTICLE IN PRESS 1572

A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

efficiency of light trap 1.0

trap efficiency

0.8

0.6

0.4

n = 1.5 n=2 n = 1.5, θc =16.45°

0.2

Fig. 4. Reduced area light trap. The area of the light trap may be chosen smaller than the area of light incidence. In the figure, the area of incidence is the lens area. A smaller light trap is advantageous when considering losses caused by imperfect reflections.

0.0 0

1

2 3 area ratio As/Ai

4

5

Fig. 2. Trapping efficiency vs. As/Ai for two values of the refractive index. The top curve is for the combination of the light trap with a photonic structure in a concentrating system as described in Section 3.3. The angular selective structure restricts the angular acceptance range to an angle of 16.451. For high area ratios high efficiencies are reached.

internal radiation 4.5 Fig. 5. Reduced size light trap. All sides As apart from the top A1 are covered with solar cells. The length of a side of the cube being a, the solar cell area is 5a2. In this case the entire top face is the input with an area of a2. Assuming a splitting of the light into three parts, we get F ¼ 5/3 and Ztrap ¼ 0.79.

4.0 n = 1.5 n=2

3.5

|int / |inc

3.0

being the input and AS1, AS2, etc. being the solar cell areas. A small light trap has also other advantages:

2.5

 Absorption losses within the trap are minimized.  Assuming perfect diffusivity, the light intensity is more

2.0 1.5

uniform.

 Because of the small volume it is possible to use an expensive,

1.0

highly refractive material for the trap.

0.5 0.0 0

1

2 3 area ratio As/Ai

4

5

Fig. 3. Internal radiation density vs. area ratio for two different refractive indices of the trap n ¼ 1.5 and 2. For larger ratios of As to Ai, the refractive index of the trap has less and less influence on the ratio Iint/Iinc.

The internal radiation density is shown in Fig. 3 against area ratio As/Ai. It shows that the refractive index becomes less important at larger area ratios.

The entire solar cell area of the cube in Fig. 5 is 5a2 with a as the dimension of the cube. If the solar spectrum is split into three parts, As/Ai ¼ 5/3 and Ztrap ¼ 0.79. So even with very small light traps high efficiency can be achieved. Similar concepts to the one given above have been proposed by Luque et al. [9] or Ortabasi [10]. In the concept of Luque a light confining cavity resembling an integrating sphere is used to reflect back the photons reflected by the solar cell. This system has already been examined closely [11,12] and high efficiencies have been predicted [8,13]. Ortabasi proposed to use a cavity into which the sunlight is concentrated. The cavity is covered on the inside with different solar cells and diffuse mirrors. In our terms, these systems would be described with a refractive index of the light trap of n ¼ 1.

2.2. Optimization of the geometric concentration light trap

2.3. The effect of initial absorption

The light trap sketched in Fig. 1 has the same area as the concentrating elements. There is no reason for the light trap to be so large. On the contrary, imperfect mirrors cause losses. This leads us to the design shown in Fig. 4. In final consequence, we end up with a light trap that is entirely covered by solar cells (with exception of the input opening). As an example see Fig. 5. The light trap is a cube with Ai

After understanding the basic principle of the light trap system, we are now able to consider initial absorption. Consider a light trap as in Fig. 5, with all the walls but Ai covered with one type of perfectly absorbing solar cells. The trapping efficiency should be Ztrap ¼ 100%, as all the incoming light is absorbed by solar cells. However, this result is not given by Eq. (1). The conclusion is that we have to take the direction of the

ARTICLE IN PRESS A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

radiation into account, as the light is not randomized immediately. Light is absorbed by the solar cells already before it is randomized and fills the trap. We call this absorption ‘‘initial absorption’’. After the initial absorption, the light is randomized. The intensity of the diffuse light inside the trap in equilibrium with Iinc is denoted Iint. We know from [5] that the diffuse radiation with the intensity Iint impinging from the inside of the trap onto the surface which is coupled out equals Iint/n2. The rest is totally internally reflected. However, this loss and the internal reflection can only occur for the light which actually reaches the surface again. Therefore, the intensity of the light from which the loss has to be calculated is reduced by the initial absorption. We define an effective reflection coefficient Rtrap as the fraction of light that actually reaches the incoupling surface after a ‘‘round trip’’ through the trap. The absorption (1Rtrap) includes light absorbed after multiple reflections and all light absorbed before the reflected light arrives again at the incoupling surface. As in Section 2.1, Rtrap applies to each frequency band separately. Rtrap depends on the geometry of the trap and the degree of spectrum splitting. It is advantageous if the bottom of the trap is diffusely reflecting (besides being absorbing for part of the spectrum) because then more light will be reflected to the vertical sides and have a chance to be absorbed. The efficiency of this light trap will now be calculated. The relevant quantities are defined in Fig. 6. Iinc is the incident radiation and the incident energy is Einc ¼ IincAi. It is assumed that the light traverses the trap surface without loss but is diffused when entering the trap. The energy reflected back towards the surface is ER ¼ Iinc Rtrap Ai

(4)

Eq. (4) can be seen as a definition of Rtrap (according to the above definition of Rtrap, it refers to the input area and not cell areas). The energy absorbed by the solar cells is Eabs ¼ Iinc ð1  Rtrap ÞAi þ Iint As

(5)

1573

Then Eesc ¼

Iint Ai Iinc Rtrap Ai ¼ n2 1 þ n2 F

and Einc  Eesc Iinc Ai  ðIinc Rtrap Ai Þ=ð1 þ n2 FÞ ¼ Iinc Ai Einc Rtrap ¼1 1 þ n2 F

Ztot ¼

(8)

We now check the validity of these relations. When Rtrap ¼ 0, then Iint ¼ 0 and Ztot ¼ 1. In this case, all radiation is absorbed in the trap as was postulated above. When Rtrap ¼ 1, then Ztot ¼ Ztrap; this is the case in which all incident light in the trap is diffuse as was assumed in Eq. (1). Normally, Rtrap ¼ 1 implies fully reflecting walls which also requires Ztrap ¼ 0. It is obvious that immediate conversion of the incoming radiation to diffuse radiation in a trap with partly absorbing walls can only be realized with scattering centres inside the trap. Therefore, the initial considerations in Section 2.2 are useful for understanding the light trap but do not fully describe the light trapping mechanism. If we compare Eq. (6) with Eq. (3), we see that it differs by the factor Rtrap as we expect from the the definition of Rtrap. In Fig. 7, Ztot is plotted vs. F for different values of Rtrap and n ¼ 1.5. This figure shows the strong influence of Rtrap on the efficiency. The dashed lower curve in Fig. 7 (R ¼ 1) represents the fully diffuse case that was also presented in Fig. 2. It is significantly below the curves with lower R.

3. The photonic light trap 3.1. The principle of the photonic light trap

The energy emitted by the trap is Eesc ¼

In this section, the second concept for light trapping is described. This concept uses a photonic structure with a distinct angular selectivity in combination with a scattering reflector. Photonic structures have recently been considered for improvement of thermophotovoltaic conversion [14]. Among the

Iint Ai n2

Setting up the energy balance and solving for Iint Einc ¼ Iinc Ai ¼ Eesc þ Eabs ¼

Iint ¼

Iint Ai þ Iinc ð1  Rtrap ÞAi þ Iint As n2

total efficiency

(6)

1.0

Iinc Rtrap ð1=n2 Þ þ F

(7)

0.8

efficiency

where F ¼ As/Ai.

Iinc Ai

Iint

AS

AS

Rtrap . Iinc

0.6

R = 0.2 R = 0.4 R = 0.6 R = 1.0 n = 0.2, θc =16.45°

0.4

0.2

0.0 AS

AS

AS

solar cells

Fig. 6. Definition of quantities used in calculation of efficiency. Iinc is the incident intensity outside the trap. It enters the trap through an incident area Ai. In the trap, a fraction of the light is initially absorbed by solar cells with an area As. The rest is reflected, so that the reflected intensity is RtrapIinc. Rtrap denotes the effective reflection of the trap. The intensity of the diffuse light inside the trap that is in equilibrium with Iinc is called Iint.

0

1

2 3 area ratio As/Ai

4

5

Fig. 7. Total efficiency including internal absorption of first pass of light for different values of trap reflectivity vs. coverage factor F ¼ As/Ai. The lower dashed curve is the case without consideration of primary absorption; the top curve is with photonic structure at the input (see Fig. 8). The angular selective structure restricts the angular acceptance range to an angle of 16.451.

ARTICLE IN PRESS 1574

A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

photonic crystals, optical elements with unique optical properties [15] exhibiting spectral and angular selectivity can be found. In Fig. 8, the principle of a light trap with an angular selective filter is sketched. A transparent light trap with refractive index n is covered with the photonic structure. The optimum photonic structure for this purpose has the following properties:

 Broad spectral bandwidth to transmit all or most of the solar 

spectrum for an angular range close to perpendicular incidence. Narrow angular selectivity to reflect all light that impinges on the filter under an angle greater a critical angle Yc.

With these properties, a light trap can function as a concentrator without other optical components. As in the preceding sections, the photonic concentrating trap can only accept direct radiation and has to track the sun. Light that impinges on the trap under perpendicular incidence can traverse the filter and enter the trap. The light is than scattered into a certain angular characteristic that is assumed to be lambertian in the following. This is realized by the application of diffuse reflectors and multiple reflections [6]. Furthermore, we assume a critical angle of Yc ¼ 11 which is compatible with the angle subtended by the sun of Yc ¼ 0.271 and with the tracking accuracy of present trackers [16]. The opening angle of the acceptance is modified by refraction upon entry into the light trap. In that way, the angular range Yc under which light can leave the trap and the filter is confined to a very narrow escape cone characterized by Yint. Rays outside this cone are reflected either by total internal reflection inside the trap or at the filter. This concept has already been applied to the increase of the path length of light in solar cells [17]. The solar cells are optimized for spectral selectivity and are covered with spectrally selective mirrors. The optical properties of the photonic light trap are now analyzed. For an idealized treatment, we first assume perfect transmission of the light trap, perfect reflectivity of the mirrored surfaces, uniformity of diffuse radiation and perfect absorption of the solar cells. The concept is for large area devices. For the clarity of the presentation, we therefore neglect the edge surfaces in our calculations. The internal escape cone angle is   sin Yc (9) Yint ¼ arcsin n

The escaping intensity is Z 2p Z Yint Iesc ¼ Ai Bint cos ðWÞ sin ðWÞ dW df ¼ pBint 0

(10)

0

where Ai is the entire surface area of the trap and Bint is the internal radiation density per steradian. We now relate Bint to the internal unidirectional radiation intensity at the boundaries of the light trap: Z 2p Z p=2 Iint ¼ Bint cos ðWÞ sin ðWÞ dW dWf ¼ pBint (11) 0

0

Therefore,   2 sin Yc AI ¼ i 2int sin ðYc Þ2 Iesc ¼ Ai Iint sin arcsin n n

(12)

The light absorbed by the solar cells is IintAs. Now we can calculate the efficiency of the photonic light trap by setting up the energy balance between incoming radiation and light absorbed by solar cells and light reemitted from the surface: Ai Iint sin ðYc Þ2 þ Iint As n2

Iinc Ai ¼

(13)

Solving for Iint, Iint ¼

Iinc

(14)

ðsinðYc Þ2 =n2 Þ þ F

with F ¼ As/Ai. The trapping efficiency is

Ztrap ¼

Iint As 1 ¼ Iinc Ai ðsin ðyc Þ2 =n2 FÞ þ 1

(15)

The efficiency depends strongly on the acceptance angle (Fig. 9). The concentration is C¼

Iint 1 ¼ Iinc ðsin ðYc Þ2 =n2 Þ þ F

(16)

High concentration can only be reached with small acceptance angles. 1/F can be interpreted as the geometric concentration (presently known angular selective filters do not yet show a spectrally independent characteristic. The development of structures

efficiency of photonic light trap 1.0

As/Ai = 0.10 As/Ai = 0.02 As/Ai = 0.03

photonic structure light trap

efficiency

0.8

0.6

0.4

0.2

diffuse reflector 0.0 0 solar cells Fig. 8. Principle of light trap with angular selective photonic structure. The photonic structure has a critical acceptance angle Yc. Light that is scattered after diffusion into an angle greater than Yc is trapped. The solar cells are optimized for spectral selectivity and are covered with spectrally selective mirrors.

20

40 60 acceptance angle (deg)

80

Fig. 9. Efficiency of photonic trap vs. opening angle Yc by restricting the acceptance angle of the photonic structure, the efficiency of the light trap is increased significantly. The maximum possible restriction is to the appearance angle of the sun of 0.271.

ARTICLE IN PRESS A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

1575

with an adjusted angular characteristic is probably the most difficult part of this concept). 3.2. More realistic systems The photonic trap does not require optical elements for concentration. Concentration occurs inside the trap. This type of trap therefore has to have a large area which means that reflection losses and absorption play a crucial role. Also, the solar cells may not be fully absorbing. All these factors will now be considered. If the reflecting areas have a reflection factor R, then the amount of light lost is Erefl ¼ Ai Iint ð1  RÞð1  FÞ There will also be absorption in the volume. The amount of light lost in a trap with thickness W and absorption coefficient a is ZZ Z 2p Z p Eabs ¼ aBint dV dO ¼ aBint WAi sin W dW df 0

0

¼ 4Iint aWAi In the case of volume absorption, we have to take the bidirectional intensity 2Iint. If the solar cell absorption is Zabs,cell, then the light absorbed by the cell is Eabs;cell ¼ Iint Zabs;cell As

Fig. 10. Cross-section of a Flatcon concentrating system. Light is concentrated by Fresnel lenses onto highly efficient solar cells. The opening angle here is Yc ¼ 16.451. Typical concentrations reached are in the order of 500.

Again we can equate the gains and the losses: Iinc Ai ¼

Ai Iint sin ðYc Þ2 þ Iint As Zabs;cell n2 þ Ai Iint ð1  RÞð1  FÞ þ 4Iint aWAi

In analogy to Eqs. (12)–(14), this results in the trapping efficiency:

Ztrap ¼

Zabs;cell F ðsin ðYc Þ

2

=n2 Þ

þ Zabs;cell F þ ð1  RÞð1  FÞ þ 4aW

(17)

3.3. Combination of geometric concentration and photonic light trap Optimal efficiency is obtained by combining both concepts described so far. In Section 2, we have implicitly assumed that the light entering the trap is coming from the full half space as it would at very high concentration. In reality, the light arriving at the trap from a concentrating lens has a limited opening angle. In Fig. 10, the situation in a real system (Flatcon) [18] is shown true to scale (the situation considered here is not compatible with secondaries). In this case, the opening angle is Yc ¼ 16.451. If a photonic structure with this acceptance angle is placed on the surface of the trap no light will be lost. On the other hand, light coming from the inside of the trap can only escape within the same angle modified by refraction. Inserting this value into Eq. (14), we obtain the upper curve in Fig. 2. Efficiency for this structure is very high. This combination has several advantages:

 The light trap is very small, so the losses due to absorption and reflective surfaces are negligible.

 Solar cell area can be of the same size as the input area, so efficiency is much higher.

 The surface area of the trap is small, so the cost of the photonic structure is tolerable. The combination light trap probably cannot have a diffusing entry aperture. Randomization of the light can be achieved by diffusing reflectors or forward scattering centres close to the surface.

The photonic system can also be analyzed with inclusion of initial absorption (Section 2.3). In order to calculate its efficiency, we modify Eq. (6) by inserting Eq. (14): ! 1 Ztot ¼ 1  Rtrap 1  (18) ðsinðyc Þ2 =n2 FÞ þ 1 The upper curve in Fig. 7 is the efficiency for a reflection of R ¼ 0.5. Even in this case, the light trap is practically lossless.

4. Solar cell technology for light trapping The advantage of light trapping is that solar cells of different band gaps can be placed side by side at the light trap. This offers several advantages:

 The solar cells can consist of different semiconductor materials.

 They do not have to be grown monolithically with very small lattice mismatch.

 More abundant materials can be used.  No tunnel diodes between the cells are needed. On the other hand, light trapping is not 100% efficient. Furthermore, the solar cells have to be equipped with selective mirrors. The selective reflection can be simplified if it is taken into account that semiconductors are transparent for sub-band-gap light. This light can be reflected by a mirror at the back of the solar cell. In Fig. 11, this is outlined for an example with three different solar cells. The largest gap cell only has a mirror at the rear of the cell. The middle cell has a mirror for the long wavelengths and an edge filter for the short wavelengths. It should be noted that the edge mirror only has to have a sharp cut off reflection but can extend far into the ultraviolet. Such mirrors are easier to realize than band reflecting mirrors. The low gap cell finally only has an edge mirror for the shorter wavelengths.

ARTICLE IN PRESS 1576

A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

Another advantage for solar cells in conjunction with light traps is that since the cells are only thin, single semiconductor cells, wrap through technology can be used for contacts. In this way, shading of the front side by contacts is avoided. Considering the selective mirrors, they do not have to be absolutely perfect regarding reflectivity and wavelength sensitivity. Reflected rays have a chance to be returned to the solar cell albeit at lower efficiency. High-energy photons absorbed by a lower gap cell are not lost but only converted at lower efficiency.

5. Stationary modules So far only light traps for concentrating systems were analyzed. They permit high concentration although deconcentration by a factor of two or three is advantageous for high trapping efficiency. These systems are best suited for climates with high direct radiation intensity. It is desirable to use the light trapping principle also for moderate climates with a high fraction of diffuse radiation. How this can be accomplished will be described now. In this paper, only stationary modules are considered. Low concentration or linear tracking systems might also be possible but will not be included

Energy Distribution [W/m2µm]

1600 1400 1200 1000 800 600 400 200 0 500

1000 1500 Wavelength [nm]

2000

1 2 3

2500

Mirrors

here. In a first approach, the transparent superstrate of the module serves as the light trap (Fig. 12). In this case, the input aperture Ai is the entire module area. The surface of the module is structured to diffuse the light. If we consider a division of the solar spectrum by two (which is most feasible for stationary systems), the geometry of Fig. 12 results. (As ¼ Ai/2) The lateral dimension of the solar cells should be of the order of the superstrate thickness or smaller. Otherwise the assumption of uniform light distribution breaks down. The ratio As/Ai ¼ F is F ¼ 0.5 and R ¼ 0.5. From Eq. (6), we derive a trapping efficiency of Ztot ¼ 0.765. This type of light trap can be improved by equipping the solar cells with a rough surface. There are however limits to this. Unlike the outer surface which is transmissive and will only randomize the incident radiation, a rough absorbing surface will lead to non-uniformity of the diffuse radiation. On the other hand, even a very modest structuring of the solar cell surface will improve efficiency noticeably. If the surface area is increased by a factor of 1.5, F ¼ 0.75, then Ztot ¼ 0.814. Better efficiency is possible with more complex geometries. Some of them are outlined here: again deconcentration is employed. The solar cell area has to be larger than the module aperture. Many geometries can be imagined to accomplish this. Fig. 13 shows two examples namely saw tooth and parallel bars. These geometries are particularly well adapted to thin film solar cells. Some aspects of the first two geometries will now be discussed. Saw tooth: If the apex angle of the saw tooth is 451, then thin film cells could be evaporated from the two sides as shown by the arrows in Fig. 13. The total efficiency is estimated as follows: As/Ai ¼ 21/2 ¼ 1.41. R can be estimated to be about R ¼ 0.4, then Ztot ¼ 0.9. Bars: For the bar As/Ai ¼ 3/2 ¼ 1.5, R could be exactly calculated for instance by ray tracing but is only estimated here. It is about R ¼ 0.25. Then efficiency Ztot ¼ 0.943. For manufacturing purposes, many bars could be placed side by side in a deposition chamber, rotated after deposition and again deposited. Finally, they could be placed in a module with the open side up. Here, just as for the saw tooth the usual thin film technologies for series connection could be used and contact could be made at the end of the bar. The structuring of the solar cell surface could of course also be applied to the geometries just mentioned.

6. Estimate of achievable conversion efficiency

Filters

Fig. 11. Solar spectrum and spectral range of selective and regular mirrors for three stages 1–3. The first solar cell absorbs light with a wavelength lower than 700 nm. Higher wavelengths are transmitted and are reflected by a mirror. The second cell absorbs light with a wavelength smaller than 900 nm. To it a filter is attached that reflects light with a wavelength smaller than 700 nm. The mirror on the backside reflects the transmitted light. The third solar cell absorbs light with a wavelength smaller than 1800 nm. To it a filter is attached that reflects all light with a wavelength lower than 900 nm. By that means a spectrum splitting is realized. The shaded areas are schematic and only give losses due to band-gap energy.

6.1. Solar cell efficiency We first consider solar cell efficiency and then look at influences of the optical system. Starting with the thermodynamic limit, we know that efficiency increases significantly with the number of solar cells i.e. the degree of spectrum splitting. Table 1 gives the thermodynamic limit of efficiency vs. number of cells [19].

Transparent superstrate

Solar cells Fig. 12. Flat panel module with half-spectrum solar cells. The solar cells can either be small squares or, more practically, elongated strips. In the figure, two kinds of solar cells are shown, splitting the spectrum into two parts. Selective mirrors or reflectors as in Fig. 11 are not shown.

ARTICLE IN PRESS A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

1577

Parallel bars Light trap

Sawtooth

Light trap

Solar cells Solar cells Fig. 13. Two versions of deconcentration in a flat plate module: (a) saw tooth and (b) parallel bars.

occurs within the trap. That means very high light intensity within the trap. Some loss factors are very severe:

Table 1 Dependence of thermodynamic efficiency limit vs. number of cells Number of cells Efficiency

1 41.1

2 54.6

3 61.9

4 65.6

5 67.9

We see that these efficiencies are very high. The biggest increase occurs when going from one cell to two cells. Those ultimate efficiencies can of course not be reached in practice. Barnett [5] estimates for a similar system practical efficiency by using derating factors. Higher band-gap cells are less derated than those with lower band gap. They estimate that for six junctions in four optical bins Z ¼ 53.5% can be achieved. Best experimental result so far is for three junctions in three optical bands Z ¼ 42.8%, estimated from separate cell measurements under the conditions of their system. With our light trap similar results can be anticipated. In order to limit the complication on the electrical side, it is suggested here that for high efficiency dual junction devices should be used for each optical band. 6.2. Efficiency of light trap for concentrators As outlined in Sections 2 and 3, trapping efficiency can be very high if optimal geometrical and optical conditions are chosen, in particular if the light trap is combined with a photonic structure. Thus, the ultimate efficiency of the system light trap-solar cells can be very close to the values given above. At this time, it is very difficult to estimate more realistic efficiencies. The following factors have to be considered: (a) Absorption within the trap. This effect can be kept small because highly transparent materials are available and the trap can be made very small. (b) The absorption of solar cells may be less than 100%. (c) Efficiency of selective mirrors. Required are no parasitic absorption and good spectral selectivity as well as good reflectivity at all incidence angles. (d) The requirement to create diffuse radiation is difficult to reconcile with selective mirrors and photonic structures. If this is not possible non-selective lambertian mirrors will have to be included. They should have very high reflectivity. The overall system efficiency is further influenced by the efficiency of the optical concentration system as in conventional systems.

(a) Absorption within the trap. When using very highly transparent material absorption could be tolerable. (b) Reflectivity of mirrors. If high concentration is desired, a large fraction of the rear surface of the trap is covered by reflectors. Reflectivity has to be extremely good to enable high efficiency. (c) Reflectivity of the photonic structure outside the escape cone. We assume total internal reflection but this has not yet been demonstrated (photonic structures can be as good as total internal reflection in principle). (d) Efficiency of selective mirrors and absorption of solar cells have already been discussed in the previous paragraph.

6.4. Stationary flat panel modules For the light trapping flat plate modules described in Section 5 the same solar cell efficiencies as discussed in Section 6.1 can be anticipated if optimal solar cells are used. Dependent on the geometry, lower efficiency of the light trap has to be taken into account. Optimal solar cells, however, may be too expensive for this application. Thin film cells are less costly and could be deposited onto the light traps. Amorphous Si is probably not capable of higher efficiency as the past development has shown. Better prospects are connected with micromorph films consisting of amorphous and microcrystalline films. Their efficiency is expected to be higher if used separately and not in a stacked cell. Even more efficiency can be foreseen for chalcogenides. The theoretical efficiency for two gaps has been estimated to be Z ¼ 36.6% [20]. In practice, Z ¼ 20% can be expected for a two gap system. Compared with concentrating systems flat plates have higher yield because they also utilise diffuse light.

7. Conclusions The light trapping principle, based on diffuse radiation inside a transparent matrix, has been shown to offer many interesting aspects. It can be used for concentrating, tracking systems but also for stationary modules. The possibility to employ solar cells consisting of different materials and band gaps offers a much wider design space and holds hope to achieve higher conversion efficiency. The present paper presents mainly the basic principle and fundamental theory. Much experimental work is probably required to explore the practical applicability of this concept. Initial experiments are under way.

6.3. Efficiency of the photonic structure light trap

Acknowledgement

The photonic structure light trap as described in Section 3 is large area and does not need optical components. Concentration

We acknowledge valuable discussions with Dr. Andreas Gombert and Dr. Benedikt Bla¨si.

ARTICLE IN PRESS 1578

A. Goetzberger et al. / Solar Energy Materials & Solar Cells 92 (2008) 1570–1578

References [1] W. Shockley, H.J. Queisser, Detailed balance limit of efficiency of p–n junction solar cells, J. Appl. Phys. 32 (3) (1961) 510–519. [2] A.W. Bett, et al., Metamorphic GaInP-GaInAs layers for photovoltaic applications, Mater. Res. Soc. Symp. Proc. 836 (2005) L6.4.1–L6.4.12. [3] A. Goetzberger, W. Greubel, Solar energy conversion with fluorescent collectors, Appl. Phys. 14 (1977) 123–139. [4] A.G. Imenes, D.R. Mills, Spectral beam splitting technology for increased conversion efficiency in solar concentrating systems: a review, Sol. Energy Mater. Sol. Cells 84 (19) (2004) 19–69. [5] A. Barnett, Milestones toward 50% efficient solar cell modules, in: Proceedings of the 22nd European Photovoltaic Solar Energy Conference, Milan, Italy, 2007. [6] E. Yablonovitch, Statistical ray optics, J. Opt. Soc. Am. 72 (7) (1982) 899–907. [7] A. Luque, Performance of front contact silicon solar cells under concentration, Prog. Photovolt. 12 (2004) 517–528. [8] A. Luque, The confinement of light in solar cells, Sol. Energy Mater. 23 (2–4) (1991) 152–163. [9] A. Luque, et al., Angle-limited cavities for silicon solar cells, in: Proceedings of the 22nd IEEE Photovoltaic Specialists Conference, Las Vegas, NV, USA, 1991, pp. 99–104. [10] U. Ortabasi, First experiences with a multi-bandgap photovoltaic cavity converter (PVCC) module for ultimate solar-to-electricity conversion efficiency,

[11]

[12] [13]

[14] [15] [16] [17] [18] [19] [20]

in: Proceedings of the 19th European Photovoltaic Solar Energy Conference, Paris, France, 2004, pp. 2256–2560. J.C. Minano, A. Luque, I. Tobias, Light-confining cavities for photovoltaic applications based on the angular-spatial limitation of the escaping beam, Appl. Opt. 31 (16) (1992) 3114–3122. A. Luque, J.C. Minano, Optical aspects in photovoltaic energy conversion, Sol. Cells 31 (3) (1991) 237–258. P.A. Davies, et al., Calculations of efficiencies achievable using gallium arsenide and silicon cells mounted together in cavities illuminated by singlestage concentration, Sol. Energy Mater. Sol. Cells 26 (3) (1992) 203–215. M. Florescu, et al., Improving solar cell efficiency using photonic band-gap materials, Sol. Energy Mater. Sol. Cells 91 (2007) 1599–1610. E. Yablonovitch, Inhibited spontaneous emission in solid-state physics and electronics, Phys. Rev. Lett. 58 (20) (1987) 2059–2062. I. Luque-Heredia, in: A. Luque, V. Andreev (Eds.), Concentrator Photovoltaics, Springer, Berlin, 2007, pp. 221–251. M. Peters, et al., Photonic crystals for the efficiency enhancement of solar cells, in: EOS Topical Meeting on Diffractive Optics, Barcelona, Spain, 2007. A.W. Bett, H. Lerchenmu¨ller, in: A. Luque, V. Andreev (Eds.), Concentrator Photovoltaics, Springer, Berlin, 2007, pp. 301–319. A.W. Bett, F. Dimroth, G. Siefer, in: A. Luque, V. Andreev (Eds.), Concentrator Photovoltaics, Springer, Berlin, 2007, pp. 67–87. H.J. Lewerenz, H. Jungblut, Photovoltaik: Grundlagen und Anwendungen, Springer, Berlin, 1995.